Longwave frequency reference receiver

Pieter-Tjerk de Boer, PA3FWM web@pa3fwm.nl

(This is an adapted version of part of an article I wrote for the Dutch amateur radio magazine Electron, May 2024.)

[EA153k] The picture shows the Schomandl EA153k 'Frequenzvergleichs-Empfänger'. Its purpose is calibrating the reference frequency of e.g. a frequency counter. The red 'bar' on top contains a ferrite antenna for 153 kHz. Via the BNC connector on the front panel a to-be-calibrated signal of 100 kHz, 1 MHz or 10 MHz can be input. The meter then indicates the phase difference between that local reference signal and the off-air 153 kHz. If the reference frequency is spot-on, the phase difference is constant and the meter's needle won't move. But if the frequency is too high or too low, the meter needle will move to the left or to the right, at a speed that depends on the frequency error.

That's a nice idea, but only if the received 153 kHz signal is precisely on frequency. The 153 kHz Deutschlandfunk transmitter in Donebach (Germany) indeed was locked to an atomic clock; but this transmitter was switched off at the end of 2014. There is still another transmitter active, near Braşov in Romania, and this receiver works fine with it, even indoors in the Netherlands. But unfortunately that transmitter is about half a Hertz too high in frequency: no problem for its audio, but it makes it useless as a frequency reference.

Principle of operation and options for conversion

The figure shows the blockdiagram of the device, drawn by myself because [at the time of writing - but see below] I did not have the original documentation. Documentation of its predecessor, the EF151, was available [1]. We see that the local reference signal is divided down to 50 kHz, 4 kHz and 2 kHz. From the 50 kHz, the third harmonic at 150 kHz is obtained and mixed with the 2 kHz to get 152 kHz. The off-air 153 kHz signal is mixed with this, giving 1 kHz. Finally, this 1 kHz is compared to the 4 kHz in a phase comparator: the meter effectively indicates how much time there is between a pulse of the 1 kHz and the next pulse of the 4 kHz.

If we'd like to convert this to a different off-air frequency, two things need to be done: converting the tuned circuits in the receiver to the new frequency, and modify the frequency synthesizer such that it gives a frequency 1 kHz above or below the new off-air frequency. And all of this preferably as simply as possible, and in the original style (namely using simple CMOS logic ICs).

There are still a number of long-wave transmitters on the air in Europe that are locked to atomic clocks: 60 kHz (England), 77.5 kHz (Germany), 162 kHz (France) and 198 kHz (England). I selected the 162 kHz. It is closest to the original 153 kHz, which simplifies the modification of the tuned circuits: most already had enough tuning range. Furthermore, it is to be expected that this transmitter will stay on the air for quite some years to come (cf. [3]).

Converting the synthesizer

But how can we generate a signal of 161 or 163 kHz, starting from 100 kHz (or 1 or 10 MHz)? We can keep the original setup as much as possible if in the last stage stage we don't mix with 2 kHz but with 11 kHz (since 150 + 11 = 161 kHz). But then we have to generate the rather odd frequency of 11 kHz. The next figure shows (at top-left) my solution.

We start with signals of 20 kHz and 2 kHz, that were already present in the existing circuit. Those are mixed in an EXOR gate, which results in the sum and difference frequencies, i.e., 18 and 22 kHz, and a bunch of weaker signals on other multiples of 2 kHz. Then we divide this signal by 2 in a flipflop, which gives 22/2 = 11 kHz.

But why does the flipflop output 11 kHz, and not (also) 18/2 = 9 kHz? That's due to the fortunate circumstance that the 2 kHz signal lags the 20 kHz signal by a little bit, due to the propagation delay of the divider chips that generate these signals. This is shown in the bottom part of the figure: the transitions of the 2 kHz signal are slightly later than those of the 20 kHz. Due to this, there's a short spike on the output of the EXOR gate, bringing the total number of pulses to 22 per millisecond; without those spikes there would have been 18 per millisecond. The spectrum at this point is not much affected by those spikes (there's not much energy in them), so in the spectrum the 18 and 22 kHz are almost equally strong. But the flipflop does respond to the spikes; its output signal clearly has 11, albeit not all equally long, pulses in 1 millisecond, which is 11 kHz (with a bit of jitter). Without the spikes we would get 9 pulses per ms, 9 kHz.

If we look more carefully, we see that the EXOR gate's output pattern repeats every 0.5 ms. As a consequence, its spectrum only has components on multiples of 2 kHz (like 18 and 22 kHz, and some others), as shown top center in the figure. The output signal of the flipflop repeats every 1 ms, so its spectrum might contain components on all multiples of 1 kHz. However, in the second half of each millisecond this signal is exactly the inverse of the first half millisecond. This symmetry suppresses all even harmonics (like in a symmetric square wave). Thus, the signal only contains odd harmonics of 1 kHz, as shown top right. And that's nice, as it prevents us from unintentionally generating a 162 kHz signal which might disturb the reception.

The 4046

At first glance the practical realisation requires two ICs, one with EXOR gates and one with flipflops. That's no problem, but using only a single IC would be nice. And that's possible, using the 4046 in an unorthodox way.

The 4046 contains a complete PLL, with two phase comparators and a VCO (voltage-controlled oscillator). One of those phase comparators is a simple EXOR gate, exactly what is needed. (The other, not used here, is a more advanced one that does not just compare phase but also frequency.) And it turns out that the VCO is built around a flipflop.

The principle schematic of the VCO is in the next figure (from [2]). It revolves around capacitor C1, connected externally and determining the oscillation frequency. The OR and NAND gates at the bottom form a flipflop, which determines whether p4 and n3, or p5 and n2 conduct. These FETs determine in which direction the current supplied by p2 (dis)charges the capacitor. For example, assume that p4 and n3 are conducting. Then the left-hand side of C1 is charged positively (red arrows) while the right-hand side is grounded. When the voltage on the left-hand side is high enough, inverters 1-4 toggle the flipflop, causing p5 and n2 to conduct so the capacitor is charged in the opposite direction (blue arrows). This continues until the right-hand side is sufficiently positive, causing the flipflop to toggle again, and the story repeats. This makes it into an oscillator, of which the frequency is determined by C1 and by the current supplied by p2. Transistors p1 and p2 together are a current mirror, so in the end it's R1 and R2 that determine how quickly C1 is charged.

For my purpose, I didn't need an oscillator, but a simple toggle flipflop to divide the 22 kHz down to 11 kHz. I achieved this by leaving out R1 and R2 (so there's no current through p2), and offering the to-be-divided frequency via two 470 pF capacitors to pins 6 and 7 (i.e., the pins normally connected to C1). This converts the oscillator into a toggle flipflop: whenever there's an input pulse, the flipflop goes to its 'other' state.

Phase modulation

With this conversion the receiver indeed works on 162 kHz. But the practical usability is rather disappointing. The reason is that the 162 kHz carrier is phase modulated by a time code, and with further phase modulation that once was designed for further data services, but was never really used for that. That phase modulation is plus and minus 57 degrees, while the meter only has a range of 90 degrees. The phase modulation is attenuated a bit by the analog 1 kHz filter following the mixer [which in fact I needed to tune to make it symmetric around 1 kHz, otherwise the phase modulation would cause cycle slips], but the needle is jittering so much that it's hard to see in which direction it is moving 'on average'.

Only during the last second of each minute there is phase modulation, which serves as the marker of the next minute in the time code. Only during this brief time the needle is steady, or moves smoothly, allowing one to actually judge the frequency accuracy of the local reference.

Old technology vs. microcontrollers

While doing this conversion, I repeatedly thought 'but this can be done so easily with a microcontroller'.

For example, generating the 161 kHz could be done by clocking a microcontroller from the local 10 MHz, and writing a program for it that contains a loop that takes exactly 10000 clock cycles, so 1 ms, which sets an output pin to high and low 161 times. The resulting 161 kHz is not pure, because the period of 161 kHz is not a nice multiple of 0.1 µs, but a little bit of jitter doesn't matter for this application: that's just a few extra signals on 1 kHz multiples above and below the desired 161 kHz, just like in the solution described earlier.

Another place where a microcontroller could be useful, is between the phase detector and the meter. The phase modulation is a time code and thus completely predictable, so the microcontroller could subtract that from the measured phase to leave the pure carrier phase. While we're at it, the microcontroller could also do the phase comparison itself: measure the time between pulses from the mixed-down carrier and the reference.

Or combine it all: clock the microcontroller from the 10 MHz reference, use an analog-to-digital converter to directly sample the received 162 kHz signal and to all further processing in software. That's how one would do it nowadays: an SDR.

Yes... it's all possible, but not much would be left of the original equipment other than the enclosure and the ferrite antenna; that would be a pity...

B.t.w., my conversion using the CMOS chip is in fact very much in style, because the device has been converted before. Until 1986 the Donebach transmitter was on 155 kHz and that's the frequency the device was designed for. When the transmitter moved to 153 kHz, and extra IC was soldered in (probably by the manufacturer), on top of an already present IC. The extra IC is the divider from 4 to 2 kHz in the blockdiagram shown earlier. Also the frontpanel was modified for this, by applying a sticker showing the updated formula for calculating the frequency deviation.

Addendum

After this article appeared in print, Ron Postma, NL14057, contacted me. He had a slightly earlier version of this equipment, the EA155k, along with all the documentation, and he kindly allowed me to scan that. The documentation (only in German) can now be found here: a brochure, the manual and description, and the schematics.

This manual and Ron's older version of the receiver show a longer history of modifications. When the equipment was first supplied, the Donebach transmitter was still on 151 kHz (for which the earlier EF151k receiver was made), but a change to 155 kHz was already planned. To cater for this, a jumper wire on the circuitboard made the receiver work on 151 kHz, and the owner could cut this to change to 155 kHz. The later modification from 155 to 153 kHz, in the form of an extra divider from 4 to 2 kHz, isn't present in Ron's receiver. However, it may still have been used on 153 kHz, as that's the image frequency when the LO is on 154 kHz.

Another interesting point is that the manual and the circuit board seem to have been made for the phase comparator running on 20 kHz rather than 4 kHz; that gives higher phase resolution. Apparently this was changed to 4 kHz at a rather early stage, witnessed both by changes in the manual and a cut trace on the circuit board.

References

[1] https://www.mikrocontroller.net/topic/464799
[2] Texas Instruments: CD4046B Phase-Locked Loop: A Versatile Building Block for Micropower Digital and Analog Applications, Application Report SCH002A, 2/2003
[3] Technische notities van PA3FWM, Electron 4/2017; online in English

Text on this page is copyright 2024, P.T. de Boer, web@pa3fwm.nl .
Republication is only allowed with my explicit permission.